Occluded Palmprint Image Recognition using Texture, Shape and SURF keypoints

In this paper we present the results of some experiments performed on recognizing occluded palmprint images. Usually in the palmprint recognition process the main region involved in classification is the region of interest (ROI). In our experiments, exactly this region is missing. We used shape features, texture features and SURF descriptors to characterize the images. The numerical experiments are computed using two datasets. Depending on the choice of parameters to find keypoints, SURF approach gave better results than shape and texture, but it has the disadvantage of being time consuming. To reduce the computing burden, a two steps method that combines shape or texture features and SURF descriptors is proposed in this paper. Experiments based on this approach lead to very good results with less computing time than the SURF-based approach. This combine procedure can be used with other methods that allow reducing the search space for the keypoint algorithm.